为快速获得高品质彩色图像,提出一种适合嵌入式机器视觉系统处理的BayerGWR白平衡算法。根据BayerCFA的结构特点,改变原GWR法输入参数。用红、绿、蓝三分量像素均值代替GWR法Gw假设公式中像素总值,用近白点三分量像素均值代替GWR和Retinex假设公式中三分量像素最大值。测试表明:该算法能够有效地减小色温差异,数据计算量小。在主频为600MHz的智能相机中,对大小为640像素×480像素的Bayer图像数据的处理运行需34ms,能够满足实际应用对实用性及实时性的要求。
The Bayer gray world-retinex(GWR) algorithm is proposed to ohtairt the high-quality color image for embedded machine vision system. The method changes the original GWR input parameters according to the structure of Bayer color filter array(CFA). Each mean value tricolor is used to replace their sum value in the gray world (GW) assumes of GWR, and the each mean value of the near white point is adopted to replace their maximum in the Retinex assumes of GWR. Test shows that the method can effectively reduce the color temperature with a low data computation and process the 640 X 480 original image data in the smart camera with 600 MHz within 34 ms. It can satisfy the practical application of practical and real-time requirements.